Seeing red in cyclic stratigraphy: Spectral noise estimation for astrochronology

Fundamental to the development of astronomical time scales is the recognition of oscillatory variability within stratigraphic data and its evaluation relative to a null “noise” hypothesis. In this study, Monte Carlo simulations are used to investigate the suitability of two commonly used noise hypotheses (the “conventional” and “robust” AR1 approaches), and the results highlight important limitations in both for cyclostratigraphic application. Perhaps most problematic, the robust AR1 method can result in inflated confidence level estimates and excessive clumping of false positives within the low frequency portion of the spectrum, especially when the underlying noise process has a high lag‐1 autocorrelation. Given typical cyclostratigraphic records, this technique will often impose “significant” eccentricity band variability, even in the case of pure AR1 noise. An alternative spectral noise estimation method is proposed to overcome these problems, which simultaneously allows for departures from the AR1 assumption, and obtains high statistical power—that is, the ability to accurately identify astronomical signals when they are present in the data. We apply the method to un‐tuned δ 18 O data from Miocene sediments of the Ceara Rise, indicating statistically significant spectral power at frequencies that are consistent with the published orbital interpretation of Weedon et al. (1997). Furthermore, evaluation of the frequency arrangement of the significant spatial bedding periods, using the average spectral misfit method for astrochronologic testing, reveals that the null hypothesis of no orbital influence can be rejected with a high degree of confidence (the 99.8% confidence level).